Visual Reconstructability as a Quality Metric for Flow Visualization

Abstract

We present a novel approach for the evaluation of 2D flow visualizations based on the visual reconstructability of the input vector fields. According to this metric, a visualization has high quality if the underlying data can be reliably reconstructed from the image. This approach provides visualization creators with a cost-effective means to assess the quality of visualization results objectively. We present a vision-based reconstruction system for the three most commonly-used visual representations of vector fields, namely streamlines, arrow glyphs, and line integral convolution. To demonstrate the use of visual reconstructability as a quality metric, we consider a selection of vector fields obtained from numerical simulations, containing typical flow features. We apply the three types of visualization to each dataset, and compare the visualization results based on their visual reconstructability of the original vector field.

What May Visualization Processes Optimize?

“…Note that G s is a reconstruction function, similar to what was discussed in [36]. In most cases, G s is only a rough approximation of the true inverse function F −1 .…”

“…Keim et al proposed a pipeline featuring two interacting parallel components for data mining models and visual data exploration respectively [40]. Jänicke et al examined several pipelines for comparative visualization, and discussed quality metrics for evaluating reconstructibility of visualization [36]. Bertini et al proposed an automated visualization pipeline driven by quality metrics [4].…”

A Metric for the Evaluation of Dense Vector Field Visualizations

“…They evaluate the quality of an image by comparing a salience map with a relevance map that is either user defined or computed from the visualization data with a salience map. Most closely related to the idea presented in this paper is the recently published work by Jänicke et al [22]. In their work, the idea of using the original vector field to evaluate the visualizations is exploited as well but preceded by visualization-specific reconstruction of the field from the image.…”

Evaluating 2D Flow Visualization Using Eye Tracking

“…These studies evaluated the effectiveness of visualizations from indirect perceptual indications and may be easily affected by other human factors. Besides, there are also studies using computational metrics, including saliency map [JC10, JWC*11] and image‐quality map [MK13, War08], to evaluate flow visualizations. These computational methods are objective but may not directly reflect users’ visual perception of flow visualization.…”